Legal claims defining the scope of protection, as filed with the USPTO.
1. A denoising method of preserving a clinically significant feature in reconstructed images, comprising the steps of: determining weight values for a reconstructed image, the reconstructed image having a time-variant time component and time-invariant space components, the time component having an initial time and a subsequent time, the reconstructed image at the initial time being defined as an initial reconstructed image while the reconstructed image at the subsequent time being defined as a subsequent reconstructed image; iteratively denoising the reconstructed image using an anisotropic diffusion filter to generate the subsequent reconstructed image as a diffused result; in each of said iterative denoising steps, adaptively weighting a difference between the initial reconstructed image and the subsequent reconstructed image according to a corresponding one of the weight values to generate an adaptively weighed result; and in each of said iterative denoising steps, adjusting the diffused result based upon the adaptively weighed result.
2. The denoising method of preserving a clinically significant feature in the reconstructed images according to claim 1 wherein said determining step determines the weight values based upon edges in the reconstructed images.
3. The denoising method of preserving a clinically significant feature in the reconstructed images according to claim 2 wherein said determining step determines the weight values for the edges using a Laplacian edge detector.
4. The denoising method of preserving a clinically significant feature in the reconstructed images according to claim 2 wherein said determining step determines the weight values for the edges using a Sobel edge detector.
5. The denoising method of preserving a clinically significant feature in the reconstructed images according to claim 2 wherein said determining step determines the weight values for the edges using a Curvature edge detector.
6. The denoising method of preserving a clinically significant feature in the reconstructed images according to claim 1 wherein the weight values are time-invariant.
7. The denoising method of preserving a clinically significant feature in the reconstructed images according to claim 1 wherein the reconstructed image at the initial time is smoothed prior to said determining step.
8. The denoising method of preserving a clinically significant feature in the reconstructed images according to claim 1 wherein the reconstructed image at the initial time is sharpened according to an equation S ( u 0 ) = ( C 2 C - 1 u 0 - 1 - C 2 C - 1 u ) , u 0 being the reconstructed image at the initial time, u being the reconstructed image at the subsequent time, C being a parameter for an unsharp filter.
9. The denoising method of preserving a clinically significant feature in the reconstructed images according to claim 1 wherein said iterative denoising steps have a predetermined time scale t, each of said iterative denoising steps having a specific Δt, a number of said iterative denoising steps being t/Δt.
10. The denoising method of preserving a clinically significant feature in the reconstructed images according to claim 9 wherein said Δt being smaller than 1/(2N) where N is a number of dimensions of the reconstructed image.
11. A denoising method of preserving a clinically significant feature in reconstructed images, comprising the steps of: determining a weight value representing a predetermined characteristic for a portion of a reconstructed image, the reconstructed image having a time component and space components, the time component having an initial time and subsequent times, the portion being specified by a set of the space components; iteratively denoising the reconstructed image at the subsequent time using an anisotropic diffusion filter to generate a current diffused result for a current iteration, the current diffused result becoming a previous diffused result at an end of the current iteration; in an instance of the current iteration, determining a difference in the portion between the reconstructed image at the initial time and the current diffused result; in the same instance of the current iteration, weighting the difference by one of the weight values corresponding to the space components to generate an adaptively weighed result, the weight value being time invariant; and in the same instance of the current iteration, adaptively adjusting the current diffused result based upon the adaptively weighed result.
12. The denoising method of preserving a clinically significant feature in reconstructed images according to claim 11 wherein said iteratively denoising step is iterated until a difference between the current diffused result and the previous diffused result is smaller than a predetermined amount.
13. The denoising method of preserving a clinically significant feature in the reconstructed images according to claim 11 wherein said determining step determines the weight values based upon edges in the reconstructed images.
14. The denoising method of preserving a clinically significant feature in the reconstructed images according to claim 13 wherein said determining step determines the weight values for the edges using a Laplacian edge detector.
15. The denoising method of preserving a clinically significant feature in the reconstructed images according to claim 13 wherein said determining step determines the weight values for the edges using a Sobel edge detector.
16. The denoising method of preserving a clinically significant feature in the reconstructed images according to claim 13 wherein said determining step determines the weight values for the edges using a Curvature edge detector.
17. The denoising method of preserving a clinically significant feature in the reconstructed images according to claim 11 wherein the reconstructed image at the initial time is smoothed prior to said determining step.
18. The denoising method of preserving a clinically significant feature in the reconstructed images according to claim 11 wherein the reconstructed image at the initial time is sharpened according to an equation S ( u 0 ) = ( C 2 C - 1 u 0 - 1 - C 2 C - 1 u ) , u 0 being the reconstructed image at the initial time, u being the reconstructed image at the subsequent time, C being a parameter for an unsharp filter.
19. The denoising method of preserving a clinically significant feature in the reconstructed images according to claim 11 wherein said iterative denoising steps have a predetermined time scale t, the current iteration having a specific Δt, a number of the current iterations being t/Δt.
20. The denoising method of preserving a clinically significant feature in the reconstructed images according to claim 19 wherein said Δt being smaller than 1/(2N) where N is a number of dimensions of the reconstructed image.
21. A denoising system for preserving a clinically significant feature in reconstructed images, comprising: a weight determining module including a memory and a processor configured for determining weight values for a reconstructed image, the reconstructed image having a time-variant time component and time-invariant space components, the time component having an initial time and a subsequent time; and an adaptively weighting anisotropic diffusion unit including a memory and a processor connected to said weight determining module and configured for iteratively denoising the reconstructed image using an anisotropic diffusion filter to generate the reconstructed image at the subsequent time during each one of iterations, said adaptively weighting anisotropic diffusion unit adaptively weighting for each of the iterations a difference between the reconstructed image at the initial time and the reconstructed image at the subsequent time according to a corresponding one of the weight values to generate an adaptively weighed result, said adaptively weighting anisotropic diffusion unit adjusting for each of the iterations the reconstructed image at the subsequent time based upon the adaptively weighed result.
22. The denoising system for preserving a clinically significant feature in the reconstructed images according to claim 21 wherein said weight determining module determines the weight values based upon edges in the reconstructed images.
23. The denoising system for preserving a clinically significant feature in the reconstructed images according to claim 22 wherein said weight determining module determines the weight values for the edges using a Laplacian edge detector.
24. The denoising system for preserving a clinically significant feature in the reconstructed images according to claim 22 wherein said weight determining module determines the weight values for the edges using a Sobel edge detector.
25. The denoising system for preserving a clinically significant feature in the reconstructed images according to claim 22 wherein said weight determining module determines the weight values for the edges using a Curvature edge detector.
26. The denoising system for preserving a clinically significant feature in the reconstructed images according to claim 21 wherein the weight values are time-invariant.
27. The denoising system for preserving a clinically significant feature in the reconstructed images according to claim 21 wherein the reconstructed image at the initial time is smoothed before said weight determining module determines the weight values.
28. The denoising system for preserving a clinically significant feature in the reconstructed images according to claim 21 wherein the reconstructed image at the initial time is sharpened according to an equation S ( u 0 ) = ( C 2 C - 1 u 0 - 1 - C 2 C - 1 u ) , u 0 being the reconstructed image at the initial time, u being the reconstructed image at the subsequent time, C being a parameter for an unsharp filter.
29. The denoising system for preserving a clinically significant feature in the reconstructed images according to claim 21 wherein said adaptively weighting anisotropic diffusion unit has a predetermined time scale t for the iterations, each of the iterations having a specific Δt, a number of the iterations being t/Δt.
30. The denoising system for preserving a clinically significant feature in the reconstructed images according to claim 29 wherein said Δt being smaller than 1/(2N) where N is a number of dimensions of the reconstructed image.
31. A denoising system for preserving a clinically significant feature in reconstructed images, comprising: a weight determining module including a memory and a processor configured for determining a weight value representing a predetermined characteristic for a portion of a reconstructed image, the reconstructed image having a time component and space components, the time component having an initial time and subsequent times, the portion being specified by a set of the space components; and an adaptively weighting anisotropic diffusion unit including a memory and a processor connected to said weight determining module and configured for iteratively denoising the reconstructed image at the subsequent time using an anisotropic diffusion filter to generate a current diffused result for a current iteration, the current diffused result becoming a previous diffused result at an end of the current iteration, said adaptively weighting anisotropic diffusion unit determining a difference in the portion between the reconstructed image at the initial time and the current diffused result in an instance of the current iteration, said adaptively weighting anisotropic diffusion unit weighting the difference by one of the weight values corresponding to the space components to generate an adaptively weighed result, the weight value being time invariant in the same instance of the current iteration, said adaptively weighting anisotropic diffusion unit adaptively adjusting the current diffused result based upon the adaptively weighed result in the same instance of the current iteration.
32. The denoising system for preserving a clinically significant feature in reconstructed images according to claim 31 wherein said adaptively weighting anisotropic diffusion unit iterates until a difference between the current diffused result and the previous diffused result is smaller than a predetermined amount.
33. The denoising system for preserving a clinically significant feature in the reconstructed images according to claim 31 wherein said weight determining module determines the weight values based upon edges in the reconstructed images.
34. The denoising system for preserving a clinically significant feature in the reconstructed images according to claim 33 wherein said weight determining module determines the weight values for the edges using a Laplacian edge detector.
35. The denoising system for preserving a clinically significant feature in the reconstructed images according to claim 33 wherein said weight determining module determines the weight values for the edges using a Sobel edge detector.
36. The denoising system for preserving a clinically significant feature in the reconstructed images according to claim 33 wherein said weight determining module determines the weight values for the edges using a Curvature edge detector.
37. The denoising system for preserving a clinically significant feature in the reconstructed images according to claim 31 wherein the reconstructed image at the initial time is smoothed before said weight determining module determines the weight values.
38. The denoising system for preserving a clinically significant feature in the reconstructed images according to claim 31 wherein the reconstructed image at the initial time is sharpened according to an equation S ( u 0 ) = ( C 2 C - 1 u 0 - 1 - C 2 C - 1 u ) , u 0 being the reconstructed image at the initial time, u being the reconstructed image at the subsequent time, C being a parameter for an unsharp filter.
39. The denoising system for preserving a clinically significant feature in the reconstructed images according to claim 31 wherein said iterative denoising steps have a predetermined time scale t for iterations, each of the iterations having a specific Δt, a number of the iterations being t/Δt.
40. The denoising system for preserving a clinically significant feature in the reconstructed images according to claim 39 wherein the Δt being smaller than 1/(2N) where N is a number of dimensions of the reconstructed image.
41. A denoising method of preserving a clinically significant feature in reconstructed images, comprising the steps of: determining weight values for a reconstructed image, the reconstructed image having a time-variant time component and time-invariant space components, the time component having an initial time and a subsequent time; iteratively denoising the reconstructed image using an anisotropic diffusion filter to generate the reconstructed image at the subsequent time as a diffused result; and in each of said iterative denoising steps, adaptively weighting a difference between the reconstructed image at the initial time and the reconstructed image at the subsequent time according to a corresponding one of the weight values to generate an adaptively weighed result, the reconstructed image at the subsequent time u(t n+1 ) being determined based upon u(t n+1 )=u(t n )+(t n+1 −t n )[∇·(D∇u(t n ))+W(u 0 −u(t n ))], wherein the weight values W is defined as W=1−e −|L(x,y,z)| and contributes to an adaptive weight, wherein L is a predetermined edge map of the reconstructed image at the initial time u 0 , as opposed to the reconstructed images u(t n ) and u(t n+1 ) at the subsequent times where t n , and t n+1 are respectively one of the subsequent times and D is a diffusion coefficient and ∇ u is local image geometry, in each of said iterative denoising steps, adjusting the reconstructed image based upon the adaptively weighed result.
42. The denoising method of preserving a clinically significant feature in the reconstructed images according to claim 41 wherein said iteratively denoising step is iterated until a difference between a current one of the diffused result and a previous one of the diffused result is smaller than a predetermined amount.
43. The denoising method of preserving a clinically significant feature in the reconstructed images according to claim 41 wherein said determining step determines the weight values based upon edges in the reconstructed images.
44. The denoising method of preserving a clinically significant feature in the reconstructed images according to claim 43 wherein said determining step determines the weight values for the edges using a Laplacian edge detector.
45. The denoising method of preserving a clinically significant feature in the reconstructed images according to claim 43 wherein said determining step determines the weight values for the edges using a Sobel edge detector.
46. The denoising method of preserving a clinically significant feature in the reconstructed images according to claim 43 wherein said determining step determines the weight values for the edges using a Curvature edge detector.
47. The denoising method of preserving a clinically significant feature in the reconstructed images according to claim 41 wherein the reconstructed image at the initial time is smoothed prior to said determining step.
48. The denoising method of preserving a clinically significant feature in the reconstructed images according to claim 41 wherein the reconstructed image at the initial time is sharpened according to an equation S ( u 0 ) = ( C 2 C - 1 u 0 - 1 - C 2 C - 1 u ) , u 0 being the reconstructed image at the initial time, u being the reconstructed image at the subsequent time, C being a parameter for an unsharp filter.
49. The denoising method of preserving a clinically significant feature in the reconstructed images according to claim 41 wherein said iterative denoising step has a predetermined time scale t, the current iteration having a specific Δt, a number of the current iterations being t/Δt.
50. The denoising method of preserving a clinically significant feature in the reconstructed images according to claim 49 wherein said Δt being smaller than 1/(2N) where N is a number of dimensions of the reconstructed image.
51. A denoising system for preserving a clinically significant feature in reconstructed images, comprising: a processing unit for reconstructing a reconstructed image from projection data; and a post-processing unit connected to said processing unit for determining weight values for the reconstructed image, the reconstructed image having a time-variant time component and time-invariant space components, the time component having an initial time and a subsequent time, said post-processing unit iteratively denoising the reconstructed image using an anisotropic diffusion filter to generate the reconstructed image at the subsequent time as a diffused result, said post-processing unit adaptively weighting a difference between the reconstructed image at the initial time and the reconstructed image at the subsequent time according to a corresponding one of the weight values to generate an adaptively weighed result, the reconstructed image at the subsequent time u(t n+1 ) being determined based upon u(t n+1 )=u(t n )+(t n+1 −t n )[∇(D∇u(t n ))+W(u 0 −u(t n ))] , wherein the weight values W is defined as W=l−e |L(x,y,z)| and contributes to an adaptive weight, wherein L is a predetermined edge map of the reconstructed image at the initial time u 0 ,as opposed to the reconstructed images u(t n ) and u(t n+1 ) at the subsequent times where t n and t n+1 are respectively one of the subsequent times and D is a diffusion coefficient and Δ u is local image geometry, said post-processing unit further adjusting the diffused result based upon the adaptively weighed result.
52. The denoising system for preserving a clinically significant feature in the reconstructed images according to claim 51 wherein said post-processing unit iteratively denoises until a difference between a current one of the diffused result and a previous one of the diffused result is smaller than a predetermined amount.
53. The denoising system for preserving a clinically significant feature in the reconstructed images according to claim 51 wherein said post-processing unit determines the weight values based upon edges in the reconstructed images.
54. The denoising system for preserving a clinically significant feature in the reconstructed images according to claim 53 wherein said post-processing unit determines the weight values for the edges using a Laplacian edge detector.
55. The denoising system for preserving a clinically significant feature in the reconstructed images according to claim 53 wherein said post-processing unit determines the weight values for the edges using a Sobel edge detector.
56. The denoising system for preserving a clinically significant feature in the reconstructed images according to claim 53 wherein said post-processing unit determines the weight values for the edges using a Curvature edge detector.
57. The denoising system for preserving a clinically significant feature in the reconstructed images according to claim 51 wherein said post-processing unit smoothes the reconstructed image at the initial time prior to determining the weight values.
58. The denoising system for preserving a clinically significant feature in the reconstructed images according to claim 51 wherein said post-processing unit sharpens the reconstructed image at the initial time according to an equation S ( u 0 ) = ( C 2 C - 1 u 0 - 1 - C 2 C - 1 u ) , u 0 being the reconstructed image at the initial time, u being the reconstructed image at the subsequent time, C being a parameter for an unsharp filter.
59. The denoising system for preserving a clinically significant feature in the reconstructed images according to claim 51 wherein said post-processing unit has a predetermined time scale t, the current iteration having a specific Δt, a number of the current iterations being t/Δt.
60. The denoising system for preserving a clinically significant feature in the reconstructed images according to claim 59 wherein said Δt being smaller than 1/(2N) where N is a number of dimensions of the reconstructed image.
61. The denoising method of preserving a clinically significant feature according to claim 1 wherein the initial time and the subsequent time are in iteration.
Unknown
January 20, 2015
Browse 5M+ US patents with plain-English claim translations and AI-generated analysis.